wangxiang1230 / SSTAP

Code for our CVPR 2021 Paper "Self-Supervised Learning for Semi-Supervised Temporal Action Proposal".
70 stars 8 forks source link

关于数据扰动 #11

Open neauniko opened 1 year ago

neauniko commented 1 year ago

您好,感谢您的两个工作。我是通过您PRN的工作了解到的数据扰动作为数据增强的操作,我能否问下在PRN中使用了那些方法来辅助数据扰动吗,因为在我使用其作为数据增强时,反而像您在SSTAP中关于TSM中操作的描述的一样,会导致性能下降。 谢谢您

wangxiang1230 commented 1 year ago

您好,感谢您的两个工作。我是通过您PRN的工作了解到的数据扰动作为数据增强的操作,我能否问下在PRN中使用了那些方法来辅助数据扰动吗,因为在我使用其作为数据增强时,反而像您在SSTAP中关于TSM中操作的描述的一样,会导致性能下降。 谢谢您

Hi, in PRN, we used data augmentation including random dropout, temporal shift, random dropout 2d and so on.